Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A mechanical noise suppression apparatus, comprising: one or more processors operable to: divide an input signal into frames of a predetermined time length; transform the frames into a frequency spectrum of a frequency domain; calculate, for each frequency of the frequency spectrum, a power ratio between the frequency spectrum of the frames of the input signal and a frequency spectrum of mechanical noise based on frequency spectrum information of the mechanical noise; acquire a correction coefficient based on a linear predictive coefficient representing a spectrum envelope of the frequency spectrum of the input signal; multiply the frequency spectrum information of the mechanical noise by the acquired correction coefficient to correct the frequency spectrum information of the mechanical noise to be used; determine, for each frequency of the frequency spectrum, a gain corresponding to the calculated power ratio from a gain function table in which set values of the gain corresponding to individual values of the power ratio are stored; multiply, for each frequency of the frequency spectrum, the frequency spectrum of the frames of the input signal by the gain to obtain a corrected frequency spectrum; convert the corrected frequency spectrum of the frames into frames of a time domain; and perform frame synthesis of the frames to obtain an output signal in which the mechanical noise is suppressed.
A mechanical noise suppression apparatus utilizes processors to reduce noise in audio signals. The process involves dividing an incoming audio signal into short time frames. Each frame is converted into its frequency spectrum. The system then calculates the power ratio between the input signal's frequency spectrum and a reference spectrum of mechanical noise. A correction coefficient, based on the input signal's spectral envelope (derived from linear predictive coding), modifies the mechanical noise spectrum. A gain value is determined for each frequency based on the power ratio, using a lookup table. The input signal's frequency spectrum is then multiplied by these gain values to create a corrected spectrum. Finally, this corrected frequency spectrum is converted back into the time domain and the frames are combined to create a noise-suppressed output signal.
2. The mechanical noise suppression apparatus according to claim 1 , wherein each of the set values of the gain stored in said gain function table is low when the power ratio is in the proximity of 0 dB and smoothly increases as the power ratio increases from the proximity of 0 dB such that a gradient thereof does not become discontinuous.
The mechanical noise suppression apparatus, as described previously, uses a gain function table where gain values are set based on the power ratio between the input signal and the mechanical noise. Specifically, the gain values stored in this table are designed to be low when the power ratio is near 0 dB (indicating similar power levels for signal and noise). These gain values then smoothly increase as the power ratio increases from 0 dB, ensuring there are no sudden jumps (discontinuities) in the gain, leading to a more natural-sounding noise reduction.
3. The mechanical noise suppression apparatus according to claim 2 , wherein each of the set values of the gain stored in said gain function table smoothly increases as the power ratio decreases from the proximity of 0 dB such that the gradient thereof does not become discontinuous.
In the mechanical noise suppression apparatus with a gain function table (as previously described), the gain values not only increase smoothly as the power ratio increases from 0 dB, but also increase smoothly as the power ratio *decreases* from 0 dB. Again, the smoothness ensures no abrupt changes in the gain, even when the noise level is slightly higher than the desired signal. This smoothness maintains audio quality when attempting to remove ambient mechanical noise.
4. The mechanical noise suppression apparatus according to claim 1 , wherein the mechanical noise appears in relation to a specific imaging operation of an imaging apparatus which has a peripheral sound recording function.
The mechanical noise suppression apparatus previously described is particularly designed to reduce mechanical noise that originates from a specific imaging operation, such as focusing or zooming, within an imaging device (like a camera) that also records audio. The noise is coupled with the imaging operations which produces a known mechanical noise profile that the method cancels.
5. A mechanical noise suppression method, comprising: dividing an input signal into frames of a predetermined time length; transforming the frames into a frequency spectrum of a frequency domain; calculating, for each frequency of the frequency spectrum, a power ratio between the frequency spectrum of the frames of the input signal and a frequency spectrum of the mechanical noise based on frequency spectrum information of the mechanical noise, wherein the frequency spectrum information of the mechanical noise is corrected by acquiring a correction coefficient based on a linear predictive coefficient representing a spectrum envelope of the frequency spectrum of the input signal, and multiplying the frequency spectrum information of the mechanical noise by the acquired correction coefficient; determining, for each frequency of the frequency spectrum, a gain corresponding to the calculated power ratio from a gain function table in which set values of the gain corresponding to individual values of the power ratio are stored; multiplying, for each frequency of the frequency spectrum, the frequency spectrum of the frames of the input signal by the gain to obtain a corrected frequency spectrum; converting the corrected frequency spectrum of the frames into frames of a time domain; and performing frame synthesis of the frames to obtain an output signal in which the mechanical noise is suppressed.
A method for suppressing mechanical noise involves these steps: First, an input audio signal is divided into frames. These frames are transformed into the frequency domain. For each frequency, the power ratio is calculated between the input signal's frequency spectrum and the mechanical noise's frequency spectrum. The mechanical noise spectrum is first corrected using a correction coefficient, which is derived from the input signal's spectral envelope (using linear predictive coding). A gain is determined for each frequency based on the calculated power ratio, using a gain function table. The input signal's frequency spectrum is then multiplied by the gain to obtain a corrected spectrum. This spectrum is converted back to the time domain and the frames are combined to create a noise-suppressed audio output.
6. A non-transitory computer-readable storage medium having stored thereon, a computer program having at least one code section, the at least one code section being executable by one or more processors for causing a computer to perform the steps comprising: dividing an input signal into frames of a predetermined time length; transforming the frames into a frequency spectrum of a frequency domain; calculating, for each frequency of the frequency spectrum, a power ratio between the frequency spectrum of the frames of the input signal and a frequency spectrum of mechanical noise based on frequency spectrum information of the mechanical noise, wherein the frequency spectrum information of the mechanical noise is corrected by acquiring a correction coefficient based on a linear predictive coefficient representing a spectrum envelope of the frequency spectrum of the input signal, and multiplying the frequency spectrum information of the mechanical noise by the acquired correction coefficient; determining, for each frequency of the frequency spectrum, a gain corresponding to the calculated power ratio from a gain function table in which set values of the gain corresponding to individual values of the power ratio are stored; multiplying, for each frequency of the frequency spectrum, the frequency spectrum of the frames by the gain to obtain a corrected frequency spectrum; converting the corrected frequency spectrum of the frames into frames of a time domain; and performing frame synthesis of the frames to obtain an output signal in which the mechanical noise is suppressed.
This describes a computer program stored on a non-transitory storage medium that, when executed by a processor, implements a mechanical noise suppression method. The method consists of dividing an input audio signal into frames; transforming these frames into the frequency domain; calculating, for each frequency, the power ratio between the input signal's frequency spectrum and the mechanical noise's frequency spectrum. Before calculating the power ratio, the mechanical noise spectrum is corrected using a coefficient derived from the input signal's spectral envelope. For each frequency, a gain is determined from a gain function table based on the calculated power ratio. The input signal's frequency spectrum is then multiplied by these gains to obtain a corrected frequency spectrum which is then transformed back into the time domain and the frames are combined to produce a noise-suppressed output audio signal.
7. An imaging apparatus having a peripheral sound recording function, the imaging apparatus comprising: one or more processors operable to: divide an input signal of peripheral sound obtained from a microphone into frames of a predetermined time length; transform the frames into a frequency spectrum of a frequency domain; calculate, for each frequency of the frequency spectrum, a power ratio between the frequency spectrum of the frames of the input signal and a frequency spectrum of mechanical noise based on frequency spectrum information of the mechanical noise, wherein the frequency spectrum information of the mechanical noise is corrected by acquiring a correction coefficient based on a linear predictive coefficient representing a spectrum envelope of the frequency spectrum of the input signal, and multiplying the frequency spectrum information of the mechanical noise by the acquired correction coefficient; determine, for each frequency of the frequency spectrum, a gain corresponding to the calculated power ratio from a gain function table in which set values of the gain corresponding to individual values of the power ratio are stored; multiply, for each frequency of the frequency spectrum, the frequency spectrum of the frames by the gain to obtain a corrected frequency spectrum; convert the corrected frequency spectrum of the frames into frames of a time domain; perform frame synthesis of the frames to obtain an output signal in which the mechanical noise is suppressed; and record the output signal.
An imaging apparatus (such as a camera) with sound recording capabilities utilizes mechanical noise suppression. The apparatus processes the audio input from a microphone by first dividing the audio into frames and transforming each frame into the frequency domain. The power ratio between the input signal's frequency spectrum and the mechanical noise's frequency spectrum is calculated for each frequency. The mechanical noise spectrum is first corrected using a correction coefficient derived from the input signal's spectral envelope. For each frequency, a gain is determined from a gain function table based on the calculated power ratio. The input signal's frequency spectrum is then multiplied by the gain to get a corrected frequency spectrum which is converted back to the time domain. The frames are combined to generate a noise-suppressed output. This final audio output is then recorded by the imaging apparatus.
8. A mechanical noise suppression apparatus, comprising: a framing section adapted to divide an input signal into frames of a predetermined time length; a Fourier transform section adapted to transform framed signals into a frequency spectrum of a frequency domain; a mechanical noise reduction section adapted to correct the frequency spectrum of the input signal based on frequency spectrum information of mechanical noise to suppress the mechanical noise; a spectrum information changing section adapted to change the frequency spectrum information of the mechanical noise to be used by said mechanical noise reduction section based on information regarding the input signal, wherein the spectrum information changing section corrects the frequency spectrum information of the mechanical noise stored in a noise table based on information regarding the input signal to change the frequency spectrum information of the mechanical noise to be used by the mechanical noise reduction section, wherein the spectrum information changing section calculates a parameter representative of a characteristic amount of peripheral sound based on the information regarding the input signal, acquires a correction coefficient based on the calculated parameter, and multiplies the frequency spectrum information of the mechanical noise stored in the noise table by the acquired correction coefficient to correct the frequency spectrum information of the mechanical noise, and wherein the parameter representative of the characteristic amount is a linear predictive coefficient representative of a spectrum envelope of the frequency spectrum of the input signal, and wherein the spectrum information changing section acquires, based on the linear predictive coefficient representative of the spectrum envelope, a correction coefficient for each frequency such that the value thereof decreases in a corresponding relationship to a mountain portion of the spectrum envelope and multiplies, for each frequency of the frequency spectrum, the frequency spectrum information of the mechanical noise by the acquired correction coefficient to correct the frequency spectrum information of the mechanical noise; an inverse Fourier transform section adapted to return the frequency spectrum corrected by said mechanical noise reduction section into framed signals of a time domain; and a frame synthesis section adapted to perform frame synthesis of the framed signals of frames obtained by said inverse Fourier transform section to obtain an output signal in which the mechanical noise is suppressed.
A mechanical noise suppression apparatus reduces noise by dividing an input signal into frames, transforming them into the frequency domain, and then correcting the frequency spectrum based on mechanical noise information. Critically, the apparatus *adjusts* the mechanical noise spectrum information based on characteristics of the input signal. Specifically, it calculates a parameter representing the input signal's characteristics (a linear predictive coefficient representing its spectral envelope), acquires a correction coefficient based on this parameter, and then multiplies the stored mechanical noise spectrum by this coefficient. The correction reduces the magnitude of the mechanical noise based on peaks in the spectrum envelope of the peripheral sound that is used for calculating the linear predictive coefficient, and then applies an inverse Fourier transform and frame synthesis to produce the final noise-suppressed output.
9. The mechanical noise suppression apparatus according to claim 8 , wherein the characteristic amount parameter is an average power of the input signal, and the one or more processors acquire, based on the average power of the input signal, a correction coefficient common to different frequencies such that the value thereof is low when the average power is high and multiplies the frequency spectrum information of the mechanical noise for each frequency by the acquired correction coefficient to correct the frequency spectrum information of the mechanical noise.
This mechanical noise suppression apparatus processes an input audio signal to reduce mechanical noise. It functions by first dividing the input signal into frames of a set duration, then transforming these frames into their frequency spectrum. To perform noise reduction, the apparatus corrects its pre-stored frequency spectrum information about the mechanical noise. Specifically, it calculates the **average power of the input signal** as a characteristic parameter. Based on this average power, the system derives a correction coefficient that is applied uniformly across all frequencies. This coefficient is inversely related to the input signal's average power, meaning its value is lower when the average power of the input signal is high. This acquired correction coefficient is then multiplied by the stored mechanical noise frequency spectrum information for each frequency. The resulting corrected mechanical noise spectrum is used to suppress noise from the input signal's frequency spectrum. Finally, the corrected frequency spectrum is converted back into time-domain frames and synthesized to produce an output signal with the mechanical noise suppressed.
10. The mechanical noise suppression apparatus according to claim 8 , further comprising a plurality of noise tables which store the frequency spectrum information of the mechanical noise which is used in a case in which the average power of the input signal is different from each other, and wherein said spectrum information changing section changes over the noise table from which the frequency spectrum information of the mechanical noise is to be read out over based on the average power of the input signal to change the frequency spectrum information of the mechanical noise to be used by said mechanical sound reduction section.
Building on the mechanical noise suppression apparatus previously described, this enhancement uses *multiple* noise tables, each storing mechanical noise frequency spectrum information for different average power levels of the input signal. The apparatus then selects the appropriate noise table to use based on the current average power of the input signal. This allows for more accurate mechanical noise suppression across a wider range of input signal power levels because the parameters can be optimized for different scenarios.
11. The mechanical noise suppression apparatus according to claim 8 , wherein the mechanical noise is generated in relation to a specific imaging operation in an imaging apparatus which has a peripheral sound recording function.
The mechanical noise suppression apparatus, as described in the preceding claims, is particularly useful when the mechanical noise is generated by specific operations within an imaging apparatus (such as a camera with a sound recording function). These operations could include zooming or focusing, which can introduce undesirable mechanical sounds into the recorded audio.
12. A mechanical noise suppression method, comprising: dividing an input signal into frames of a predetermined time length; transforming the frames into a frequency spectrum of a frequency domain; correcting the frequency spectrum of the frames of the input signal based on frequency spectrum information of mechanical noise to suppress the mechanical noise, wherein the frequency spectrum information of the mechanical noise to be used is corrected by: calculating a parameter representative of a characteristic amount of peripheral sound based on information regarding the input signal, wherein the parameter representative of the characteristic amount is a linear predictive coefficient representing a spectrum envelope of the frequency spectrum of the input signal; acquiring a correction coefficient for each frequency based on the linear predictive coefficient representing the spectrum envelope of the frequency spectrum of the input signal such that a value thereof decreases in a corresponding relationship to a peak curve portion of the spectrum envelope; and multiplying, for each frequency of the frequency spectrum, the frequency spectrum information of the mechanical noise by the acquired correction coefficient to correct the frequency spectrum information of the mechanical noise; converting the corrected frequency spectrum of the frames into frames of a time domain; and performing frame synthesis of the frames to obtain an output signal in which the mechanical noise is suppressed.
A mechanical noise suppression method involves: dividing an input signal into frames, transforming those frames into the frequency domain, and correcting the frequency spectrum of the input signal based on a mechanical noise spectrum to suppress noise. To improve accuracy, the mechanical noise spectrum is dynamically corrected by calculating a characteristic parameter of the input sound based on the input signal, which is a linear predictive coefficient representing a spectrum envelope of the frequency spectrum of the input signal. A frequency-dependent correction coefficient, based on this parameter, is then applied to the mechanical noise spectrum, reducing the mechanical noise based on peaks in the input sounds spectrum. After noise reduction, the corrected frequency spectrum of the frames is converted back to the time domain, and frame synthesis is performed to generate the noise-suppressed output signal.
13. A non-transitory computer-readable storage medium having stored thereon, a computer program having at least one code section, the at least one code section being executable by one or more processors for causing a computer to perform the steps comprising: dividing an input signal into frames of a predetermined time length; transforming the frames into a frequency spectrum of a frequency domain; correcting the frequency spectrum of the frames of the input signal based on frequency spectrum information of mechanical noise to suppress the mechanical noise, wherein the frequency spectrum information of the mechanical noise to be used is corrected by: calculating a parameter representative of a characteristic amount of peripheral sound based on information regarding the input signal, wherein the parameter representative of the characteristic amount is a linear predictive coefficient representing a spectrum envelope of the frequency spectrum of the input signal; acquiring a correction coefficient for each frequency based on the linear predictive coefficient representing the spectrum envelope of the frequency spectrum of the input signal such that a value thereof decreases in a corresponding relationship to a mountain portion of the spectrum envelope; and multiplying, for each frequency of the frequency spectrum, the frequency spectrum information of the mechanical noise by the acquired correction coefficient to correct the frequency spectrum information of the mechanical noise; converting the corrected frequency spectrum of the frames into frames of a time domain; and performing frame synthesis of the frames to obtain an output signal in which the mechanical noise is suppressed.
A non-transitory computer-readable storage medium stores a program that implements mechanical noise suppression. When executed, the program divides an audio signal into frames and transforms them into the frequency domain. The program then corrects the frequency spectrum by first modifying a stored mechanical noise spectrum. This modification involves calculating a linear predictive coefficient to represent the spectrum envelope of the frequency spectrum of the input signal. A correction coefficient for each frequency is acquired such that the value of the coefficient decreases based on a peak of the spectrum envelope. The mechanical noise is then corrected by multiplying the coefficients for each frequency of the frequency spectrum by the acquired correction coefficient. The program then transforms the corrected frequency spectrum back to the time domain, performs frame synthesis, and outputs the noise-suppressed signal.
14. An imaging apparatus having a peripheral sound recording function, the imaging apparatus comprising: one or more processors operable to: divide an input signal of peripheral sound obtained by sound collection by a microphone into frames of a predetermined time length; transform the frames into a frequency spectrum of a frequency domain; correct the frequency spectrum of the frames of the input signal based on frequency spectrum information of mechanical noise to suppress the mechanical noise, wherein the frequency spectrum information of the mechanical noise to be used is corrected by: calculating a parameter representative of a characteristic amount of peripheral sound based on information regarding the input signal, wherein the parameter representative of the characteristic amount is a linear predictive coefficient representing a spectrum envelope of the frequency spectrum of the input signal; acquiring a correction coefficient for each frequency based on the linear predictive coefficient representing the spectrum envelope of the frequency spectrum of the input signal such that a value thereof decreases in a corresponding relationship to a peak curve portion of the spectrum envelope; and multiplying, for each frequency of the frequency spectrum, the frequency spectrum information of the mechanical noise by the acquired correction coefficient to correct the frequency spectrum information of the mechanical noise; convert the corrected frequency spectrum of the frames into frames of a time domain; perform frame synthesis of the frames to obtain an output signal in which the mechanical noise is suppressed; and record the output signal.
An imaging device (like a camera) with audio recording capability utilizes mechanical noise suppression. Audio captured by a microphone is divided into frames, which are then transformed into the frequency domain. The frequency spectrum is then corrected to remove the mechanical noise. Specifically, the mechanical noise spectrum is corrected based on the linear predictive coefficients (LPC) of the sound that is recorded. Based on the LPC, a correction coefficient for each frequency is acquired such that the value of the coefficient decreases based on the peaks of the spectrum envelope. This coefficient is then applied to modify the mechanical noise spectrum before subtracting the noise spectrum. Finally, the corrected frequency spectrum is transformed back to the time domain, synthesized, and recorded.
15. The mechanical noise suppression method according to claim 12 , wherein the frequency spectrum information of the mechanical noise to be used is corrected based on power of the input signal.
In the mechanical noise suppression method described earlier, the correction of the mechanical noise frequency spectrum is also based on the power of the input signal. This means the amount of noise reduction applied is adjusted based on how loud the input signal is. Louder signals may have less noise reduction applied to them, because mechanical noise will be easier to filter out.
16. The mechanical noise suppression apparatus according to claim 1 , wherein the one or more processors are operable to selectively output one of: the frequency spectrum of the frames of the input signal or the corrected frequency spectrum before conversion into time domain.
The mechanical noise suppression apparatus, described earlier, has the ability to selectively output either the original, uncorrected frequency spectrum of the input signal or the corrected frequency spectrum before it is transformed back into the time domain. This provides flexibility for debugging or further processing of the signal at different stages of the noise reduction pipeline.
17. The mechanical noise suppression apparatus according to claim 1 , wherein the one or more processors are operable to store the frequency spectrum information of the mechanical noise corresponding to zooming operation in a plurality of directions.
The mechanical noise suppression apparatus, described earlier, stores different frequency spectrum information for mechanical noise corresponding to different zooming operations. This enables the noise reduction algorithm to be tailored to specific mechanical noises generated by zooming in different directions (e.g., zooming in versus zooming out), leading to more effective noise suppression.
18. The mechanical noise suppression apparatus according to claim 1 , wherein the one or more processors are operable to calculate the correction coefficient by: calculating an average spectrum for a predetermined time based on the frequency spectrum of the input signal; calculating an average spectrum envelope from the average spectrum; and calculating the correction coefficient from the average spectrum envelope.
In the mechanical noise suppression apparatus previously described, the correction coefficient used to adjust the mechanical noise spectrum is calculated as follows: First, an average spectrum is calculated over a specified period of time based on the frequency spectrum of the input signal. Then, an average spectrum envelope is derived from this average spectrum. Finally, the correction coefficient is calculated from this average spectrum envelope. This method reduces the impact of sudden spectral fluctuations.
Unknown
December 16, 2014
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.